Preclinical models of shock and sepsis: what can they tell us?

The goal of translational research is to transform biologic knowledge into new treatments for human disease. Although preclinical models replicate some of the features of the disease process modeled, they invariably fail to reproduce the complexity of human illness, and by their very experimental nature, they are readily manipulated to maximize evidence of efficacy. The result is that successful translation from preclinical models to clinically effective therapy is uncommon, and that clinical trials are often undertaken without a comprehensive and realistic preclinical portfolio of studies to optimize their design. The lethal and morbid human conditions of sepsis and shock are attractive targets for new therapies and enormously challenging processes to translate because they entail considerable clinical heterogeneity, require emergent effective intervention, and are shaped not only by the initial insult, but by approaches to subsequent resuscitation and support. A colloquium jointly sponsored by the Shock Society and the International Sepsis Forum in June 2004 addressed the challenges of translational research in shock and sepsis. Through a comprehensive review of a broad variety of model approaches, and vigorous debate about the merits of differing strategies, a series of common themes emerged. We concluded that there is no single ideal model of shock or sepsis, but rather a large number of complementary models that recapitulate some discrete features of the disorders while minimizing others. Consequently, successful preclinical investigation mandates the use of a panel of preclinical studies consciously designed to address specific questions of relevance to the clinical setting. A corollary of this conclusion is that preclinical studies can shape concepts of disease and can be used to refine decisions regarding optimal patient populations for therapeutic interventional trials. We further recognized that the design and reporting of preclinical studies is highly variable, thereby limiting effective data interpretation and integration between studies. Hence, greater model standardization would aid in interpreting data and in pooling results into systematic data syntheses: such efforts should be promoted and undertaken.